Github user hyunsik commented on the pull request:
https://github.com/apache/tajo/pull/200#issuecomment-59323696
Your patch is a very nice starting point for histogram in Apache Tajo. The
patch looks good to me. Additionally, I leave some comments.
As you mentioned, precision is less important in sampling approaches. In
many cases, using double may be a right solution. But, when it comes to value
range representation, Double still has limitation.
in many cases, Text value can be easily longer than the representation
range of DOUBLE or LONG. I already fixed many bugs of sort bugs caused by long
text value (up to 256 bytes). For ETL workloads, it is usual. Also, can Double
represent an entire Long range? We need to check it. In my opinion, we need to
use four value types: Long, Double, Text, and Byte [].
Also, I'll start other discussions in TAJO-1091
(https://issues.apache.org/jira/browse/TAJO-1091).
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
